Negative binomial linear regression model
in [SAS]
|
From: Wensui Liu on 27 Dec 2009 21:36 irin, below are my answers to your questions -Am I right trying to use this model to answer the question above? first of all, you need to test if it is necessary to do NB model instead a simple Poisson Model. Auxiliary OLS regression (Cameron, 1996) is a good tool to check such necessity. Secondly, if there is over-dispersion, you need to check the potential cause if it is due to excessive zeroes. If there are excessive zeroes, a more complex model such as ZIP/Hurdle might be better. Again, a good way to check if such extra complexity is needed is to do a vuong test. -I assume that Proc Genmod can be used on this model -What others Procs can be used? for NB model, genmod is more than enough. besides, other procs such as nlmixed, nlin, model, glimmix, countreg can be used to estimate NB model as well. On Fri, Dec 18, 2009 at 9:42 AM, Irin later <irinfigvam(a)yahoo.com> wrote: > > I am trying to construct a model for the task. This is about the factors influencing patients� visits (urgent and not) rates of ED in terms of parents education and residing factors. In the literature I found out that Negative binomial linear regression model might be used to determine the association... > > -Am I right trying to use this model to answer the question above? > -I assume that Proc Genmod can be used on this model... > -What others Procs can be used? > > Thank you in advance, > > Irin > > > -- ============================== WenSui Liu Blog : statcompute.spaces.live.com Tough Times Never Last. But Tough People Do. - Robert Schuller ==============================
|
Pages: 1 Prev: Discriminant Analysis in SAS XXXX Next: Discriminant Analysis in SAS XXXX |